基于Landsat TM与MODIS缨帽变换分量的时空数据融合方法研究
发布时间：2018-12-17 02:11
【摘要】：Landsat TM数据的高空间分辨率及多光谱特性使得其在多领域得到广泛应用,但是较长的重访周期以及云的影响导致实际可用的数据较少,极大限制了其在时序分析方面的应用。反之,MODIS数据的高时间分辨率更适用于时序分析,但MODIS250米～1000米的空间分辨率具有较少的空间细节信息,更适用于空间大尺度范围的研究。 而基于Landsat TM与MODIS数据的时空数据融合方法将TM数据的高空间分辨率与MODIS数据的高时间分辨率有效地融合在一起获得新的数据,以满足在较高的空间分辨率上进行时序变化研究。 STARFM(Spatial and Temporal Adaptive Reflectance Fusion Model)是目前应用较多、精度较高、基于反射率数据的时空融合模型之一。本文通过调整原算法中参数值大小,完成了缨帽变换分量的数据融合。通过对比分析确定了数据融合最佳的参数组合。利用绿度植被指数的周期性、短时间内的渐变性特征,引入时间权重系数,提出了针对GVI数据的时空融合模型,提高了融合影像质量。 本研究的主要贡献包括： (1)将STARFM中移动窗口、分类数、影像不确定性值、距离权重常数设置为一系列不同值,利用Landsat TM与MODIS的缨帽变换分量数据获取相应融合影像,并与实际获取的Landsat TM缨帽变换分量影像对比,结果显示：移动窗口的增大以及分类数的调整有助于融合质量的提高,而距离权重常数的变化基本不会对融合结果质量造成影响,当MODIS与TM影像不确定性值为非0值时,对结果影响较小。所以原算法中参数值的调整,对结果会产生一定的影响,但对融合结果质量的改善程度有限。 (2)本文以2007年的实际获取Landsat TM数据及相应时间的MODIS数据为例,探讨了输入影像获取时间对融合结果的影响：1)融合影像时间与输入影像获取时间相差时间越长,精度越低；2)输入影像中植被处于生长高峰期时,绿度植被指数融合影像精度相对较高,但随时间推移,植被生长状况发生显著变化时,精度会明显降低。 (3)本文在原算法基础上假设绿度植被指数在短期内呈均匀变化,提出针对GVI的GSTARFM(GVI STARFM)。GSTARFM基于两个时刻的输入影像,在相似像元选取上采用两个时刻6个缨帽分量,引入时间权重系数,使GVI融合结果得到提高。 (4)GVI时序融合影像能够显示植被生长的基本特征。植被生长、高峰以及衰落在趋势变化曲线上表现明显,且峰值大小排序以及出现时间与实地调查结果相符,表明GSTARFM的有效性。
[Abstract]:Landsat TM data is widely used in many fields due to its high spatial resolution and multispectral characteristics. However, the long period of revisiting and the influence of cloud result in less available data, which greatly limits its application in time series analysis. On the contrary, the high temporal resolution of MODIS data is more suitable for time series analysis, but the spatial resolution of MODIS250 meters to 1000 meters has less spatial detail information and is more suitable for the study of large scale spatial range. The spatio-temporal data fusion method based on Landsat TM and MODIS data can effectively fuse the high spatial resolution of TM data with the high temporal resolution of MODIS data to obtain new data to satisfy the time series change research on higher spatial resolution. STARFM (Spatial and Temporal Adaptive Reflectance Fusion Model) is one of the spatiotemporal fusion models based on reflectivity data. In this paper, the data fusion of the tasseled hat transform component is completed by adjusting the value of the parameters in the original algorithm. Through comparative analysis, the best parameter combination of data fusion is determined. Based on the periodicity of green vegetation index and the characteristics of gradual change in a short period of time, the temporal weight coefficient is introduced, and a spatio-temporal fusion model for GVI data is proposed, which improves the quality of fusion image. The main contributions of this study are as follows: (1) the moving window, the classification number, the image uncertainty and the distance weight constant in STARFM are set to a series of different values, and the corresponding fusion image is obtained by using the tassel transform component data of Landsat TM and MODIS. Compared with the actual Landsat TM tasseled cap transform image, the results show that the increase of moving window and the adjustment of classification number are helpful to the improvement of fusion quality, while the change of distance weight constant has little effect on the quality of fusion result. When the uncertain value of MODIS and TM images is non-zero, the effect on the results is small. Therefore, the adjustment of the parameters in the original algorithm will have a certain effect on the results, but the quality of the fusion results will be improved to a limited extent. (2) taking the actual Landsat TM data and the corresponding time MODIS data obtained in 2007 as an example, the paper discusses the influence of input image acquisition time on the fusion results: 1) the longer the difference between the fusion image time and the input image acquisition time, the longer the time difference between the fusion image acquisition time and the input image acquisition time; The lower the precision; 2) in the input image, the accuracy of green-degree vegetation index fusion image is relatively high when the vegetation is in the peak growth period, but with time, the precision will decrease obviously when the vegetation growth status changes significantly. (3) on the basis of the original algorithm, the green vegetation index is assumed to change uniformly in a short time, and the GSTARFM (GVI STARFM). GSTARFM for GVI is based on the input image of two times, and the two time and six tasseled cap components are used in the selection of similar pixels. The time weight coefficient is introduced to improve the fusion result of GVI. (4) GVI temporal fusion images can show the basic characteristics of vegetation growth. The vegetation growth, peak and fading are obvious on the trend curve, and the ranking of peak value and the time of occurrence are consistent with the results of field investigation, which shows the validity of GSTARFM.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP202;P237
本文编号：2383466
[Abstract]:Landsat TM data is widely used in many fields due to its high spatial resolution and multispectral characteristics. However, the long period of revisiting and the influence of cloud result in less available data, which greatly limits its application in time series analysis. On the contrary, the high temporal resolution of MODIS data is more suitable for time series analysis, but the spatial resolution of MODIS250 meters to 1000 meters has less spatial detail information and is more suitable for the study of large scale spatial range. The spatio-temporal data fusion method based on Landsat TM and MODIS data can effectively fuse the high spatial resolution of TM data with the high temporal resolution of MODIS data to obtain new data to satisfy the time series change research on higher spatial resolution. STARFM (Spatial and Temporal Adaptive Reflectance Fusion Model) is one of the spatiotemporal fusion models based on reflectivity data. In this paper, the data fusion of the tasseled hat transform component is completed by adjusting the value of the parameters in the original algorithm. Through comparative analysis, the best parameter combination of data fusion is determined. Based on the periodicity of green vegetation index and the characteristics of gradual change in a short period of time, the temporal weight coefficient is introduced, and a spatio-temporal fusion model for GVI data is proposed, which improves the quality of fusion image. The main contributions of this study are as follows: (1) the moving window, the classification number, the image uncertainty and the distance weight constant in STARFM are set to a series of different values, and the corresponding fusion image is obtained by using the tassel transform component data of Landsat TM and MODIS. Compared with the actual Landsat TM tasseled cap transform image, the results show that the increase of moving window and the adjustment of classification number are helpful to the improvement of fusion quality, while the change of distance weight constant has little effect on the quality of fusion result. When the uncertain value of MODIS and TM images is non-zero, the effect on the results is small. Therefore, the adjustment of the parameters in the original algorithm will have a certain effect on the results, but the quality of the fusion results will be improved to a limited extent. (2) taking the actual Landsat TM data and the corresponding time MODIS data obtained in 2007 as an example, the paper discusses the influence of input image acquisition time on the fusion results: 1) the longer the difference between the fusion image time and the input image acquisition time, the longer the time difference between the fusion image acquisition time and the input image acquisition time; The lower the precision; 2) in the input image, the accuracy of green-degree vegetation index fusion image is relatively high when the vegetation is in the peak growth period, but with time, the precision will decrease obviously when the vegetation growth status changes significantly. (3) on the basis of the original algorithm, the green vegetation index is assumed to change uniformly in a short time, and the GSTARFM (GVI STARFM). GSTARFM for GVI is based on the input image of two times, and the two time and six tasseled cap components are used in the selection of similar pixels. The time weight coefficient is introduced to improve the fusion result of GVI. (4) GVI temporal fusion images can show the basic characteristics of vegetation growth. The vegetation growth, peak and fading are obvious on the trend curve, and the ranking of peak value and the time of occurrence are consistent with the results of field investigation, which shows the validity of GSTARFM.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP202;P237
【相似文献】
相关期刊论文 前1条
1 孙洁娣;靳世久;孙健;;基于多地震波传感器数据融合的管道安全监测预警系统[J];石油学报;2009年03期
相关硕士学位论文 前3条
1 张鹏;基于Landsat TM与MODIS缨帽变换分量的时空数据融合方法研究[D];兰州大学;2014年
2 田林;海水透明度的卫星遥感反演及其多传感器融合方法[D];中国海洋大学;2013年
3 马国胜;基于多传感器融合技术的瓦斯监控系统实现[D];武汉理工大学;2010年
，本文编号：2383466
本文链接：https://www.wllwen.com/kejilunwen/dizhicehuilunwen/2383466.html
